Detection of soil erosion in complex terrain and steep slopes has always been a challenge. The 3D point clouds achieved by the Unmanned Aerial Vehicle-Structure from Motion (UAV-SfM) technology provides an efficient and cost-effective method for obtaining large-scale terrain data, making it an important data source for monitoring land surface changes. However, there is a lack of comprehensive research on UAV-SfM terrain change monitoring algorithms, limiting its application in the study of soil erosion and sediment transport processes. This study assessed the accuracy of four commonly used geomorphic change detection algorithms in the Pisha sandstone area of the Loess Plateau, including Digital Elevation Model of Difference (DoD), Cloud to Cloud (C2C), Cloud to Mesh (C2M), and Multiscale Model to Model Cloud Comparison (M3C2). . Point cloud data employed to operate the four algorithms were produced using the SfM technique based on images acquired by UAV between July 2022 and March 2023. The impact of point density changes in the accuracy of the employed algorithms was also investigated. Results showed that all four algorithms were capable of effectively monitoring large surface changes. Among them, the M3C2 algorithm performed the best with the highest accuracy (R2 = 0.953, p <0.01) and the lowest error (MAE = 0.0161m, MRE = 3.37%, RMSE = 0.0194m), followed by the C2M algorithm. The DoD algorithm was only suitable for flat areas and yielded overestimated results for steep sloping areas. The M3C2 and C2C algorithms were sensitive to point cloud density, while the C2M and DoD algorithms were lesssensitive. The study provided a useful reference for the selection of erosion monitoring methods for the Pisha sandstone areas.